Component failure risk monitoring and early warning method and related equipment

Abstract

The invention provides a component failure risk monitoring and early warning method and related equipment, classification of performance parameter test data sets of components is completed based on hierarchical clustering, variables with large correlation are clustered into a class cluster, and a plurality of clustering results are obtained. Modeling each clustering result based on a Gaussian mixture model, calculating an envelope analysis statistical magnitude according to the definition of the statistical magnitude after the modeling is completed, determining a target monitoring statistical magnitude by integrating the statistical magnitude values of all the clustering results, determining a target envelope baseline based on the target monitoring statistical magnitude, and testing data of the current performance parameters of the component sample, according to the comparison condition of the corresponding current statistical magnitude and the target envelope baseline, the existing failure risk type is determined and early warning is carried out, parameter tracing can be carried out on the component with the first type of failure risk, comprehensive monitoring of the mean value, the variance value and the correlation coefficient between the variables of the multiple variables is realized, the component with potential risk can be found in time, and the reliability of the component is improved. And the quality reliability of the component is improved.

Description

Technical Field

[0001] The present application relates to the technical field of electronic equipment monitoring, and in particular to a component failure risk monitoring and early warning method and related equipment. Background Art

[0002] With the development of the electronics industry, integrated circuit products are being used more and more widely in various fields. During the design and development of electronic products, there may be problems with component failure. Therefore, in order to ensure the development quality of electronic products, it is necessary to monitor the risk of component failure.

[0003] Based on the above situation, the component failure risk monitoring and early warning method used in the prior art cannot timely and effectively discover the potential failure risks of components. Summary of the invention

[0004] In view of this, the purpose of this application is to propose a component failure risk monitoring and early warning method and related equipment to solve the above-mentioned technical problems.

[0005] Based on the above objectives, the first aspect of the present application provides a component failure risk monitoring and early warning method, comprising:

[0006] Obtain performance parameter test data sets of components under normal conditions;

[0007] Performing hierarchical clustering on the performance parameter test data set to obtain multiple clustering results;

[0008] Modeling each clustering result based on the Gaussian mixture model to obtain a modeling model, and determining the Gaussian mixture model parameters of the modeling model through the maximum expectation algorithm;

[0009] Determine a statistical value corresponding to each clustering result using a modeling model with Gaussian mixture model parameters, determine a target monitoring statistic by integrating all statistical values, and determine a target degree of freedom based on a variable of the target monitoring statistic;

[0010] Searching for a target envelope baseline corresponding to the target degree of freedom and a preset confidence level from a preset database, wherein the database includes envelope baselines corresponding to different degrees of freedom and confidence levels;

[0011] Acquire current performance parameter test data of the component sample, and determine current statistics corresponding to the current performance parameter test data;

[0012] In response to the current statistic being greater than the target envelope baseline, determining that the component has a first type of failure risk, and issuing an early warning using a first early warning rule preset corresponding to the first type of failure risk; or,

[0013] In response to the current statistic being less than or equal to the target envelope baseline, it is determined that the component has a second type of failure risk, and a warning is issued using a second warning rule preset corresponding to the second type of failure risk.

[0014] Optionally, the current statistic includes a plurality of current sub-statistics respectively corresponding to current performance parameter test data;

[0015] In response to the current statistic being greater than the target envelope baseline, after determining that the component has a first type of failure risk, the method includes:

[0016] For each of the sub-statistics described the following operations are performed:

[0017] Determining a chi-square distribution density function satisfied by the sub-statistic;

[0018] Integrating the chi-square distribution density function to obtain the risk probability of the current performance parameter test data corresponding to the sub-statistic;

[0019] Sort the risk probabilities of the current performance parameter test data according to the preset sorting rules to obtain the sorting results;

[0020] A predetermined number of current performance parameter test data are selected from the sorting results as failure risk parameters of the components.

[0021] Optionally, the performance parameter test data set includes a plurality of performance parameter test data;

[0022] The hierarchical clustering is performed on the performance parameter test data set to obtain multiple clustering results, including:

[0023] Taking each performance parameter test data as an initial cluster, performing Pearson correlation coefficient processing on each of the multiple initial clusters to obtain multiple first Pearson correlation coefficients;

[0024] Counting the largest first Pearson correlation coefficient among the multiple first Pearson correlation coefficients, merging two initial clusters corresponding to the largest first Pearson correlation coefficient to obtain a first merged cluster, and using the first merged cluster as a new cluster;

[0025] A cluster merging process is performed on each of the other clusters in the multiple initial clusters except the two clusters merged into the new cluster and the new cluster. The cluster merging process is as follows:

[0026] Performing Pearson correlation coefficient processing on each of the other clusters and the new clusters, to obtain a plurality of second Pearson correlation coefficients;

[0027] The largest second Pearson correlation coefficient among the second Pearson correlation coefficients is counted, and the two clusters corresponding to the largest second Pearson correlation coefficient are merged to obtain a second merged cluster. The second merged cluster is used as the new cluster, and the cluster merging process is repeated until all initial clusters are merged into the same cluster or a preset clustering termination condition is reached to obtain a nested clustering tree, and the multiple clustering results are obtained based on the nested clustering tree.

[0028] Optionally, determining the Gaussian mixture model parameters of the modeling model by using a maximum expectation algorithm includes:

[0029] For each clustering result, the following operations are performed:

[0030] Initialize the mean, variance and prior probability of each Gaussian component;

[0031] Determine, in an initialization state, a first posterior probability that each performance parameter test data in the clustering result belongs to each Gaussian component;

[0032] Based on the first posterior probability, the mean, variance value and prior probability of each Gaussian component are updated by maximizing the likelihood function;

[0033] The updated mean of each Gaussian component is used as the target mean, the updated variance value of each Gaussian component is used as the target variance value, and the updated prior probability of each Gaussian component is used as the target prior probability. An update iteration process is performed for the target mean, target variance value, and target prior probability corresponding to each Gaussian component. The update iteration process is as follows:

[0034] Determine in an updated state the second posterior probability that each performance parameter test data in the clustering result belongs to each Gaussian component;

[0035] Based on the second posterior probability, the mean, variance value and prior probability of each Gaussian component are updated by maximizing the likelihood function, the updated mean of each Gaussian component is used as the target mean, the updated variance value of each Gaussian component is used as the target variance value, and the updated prior probability of each Gaussian component is used as the target prior probability, and the update iteration process is repeatedly performed for the target mean, target variance value and target prior probability corresponding to each Gaussian component until a preset update iteration termination condition is reached, so as to obtain the final mean, final variance value and final prior probability corresponding to each Gaussian component;

[0036] The final mean, final variance value and final prior probability corresponding to each Gaussian component are used as the Gaussian mixture model parameters.

[0037] Optionally, the modeling model with Gaussian mixture model parameters is obtained by the following process:

[0038] For each clustering result, the following operations are performed:

[0039] For each Gaussian component, determining a target Gaussian component parameter based on the corresponding final mean value and final variance value;

[0040] Based on the parameters of each target Gaussian component, the corresponding final prior probability is used for integration to obtain the modeling model with Gaussian mixture model parameters.

[0041] Optionally, the modeling model with Gaussian mixture model parameters includes multiple Gaussian components, and each clustering result includes multiple performance parameter test data;

[0042] The method of using a modeling model with Gaussian mixture model parameters to determine a statistical value corresponding to each clustering result and combining all statistical values ​​to determine a target monitoring statistic includes:

[0043] Any performance parameter test data among the plurality of performance parameter test data is used as target performance parameter test data, and the following operations are performed for each target performance parameter test data in the clustering result:

[0044] Determine the distance between the target performance parameter test data and the center of each Gaussian component, and sum each distance to obtain an initial statistic corresponding to the target performance parameter test data;

[0045] The initial statistics corresponding to the target performance parameter test data in each clustering result are summed to obtain the target monitoring statistics.

[0046] Optionally, the step of obtaining a performance parameter test data set of a component in a normal state includes:

[0047] Acquire a plurality of initial performance parameter test data of the components in a normal state;

[0048] The multiple initial performance parameter test data are normalized to obtain the performance parameter test data set.

[0049] Based on the same inventive concept, the second aspect of the present application provides a component failure risk monitoring and early warning device, comprising:

[0050] A first acquisition module is configured to acquire a performance parameter test data set of a component in a normal state;

[0051] A clustering module is configured to perform hierarchical clustering on the performance parameter test data set to obtain multiple clustering results;

[0052] A parameter determination module is configured to model each clustering result based on a Gaussian mixture model to obtain a modeling model, and determine Gaussian mixture model parameters of the modeling model by a maximum expectation algorithm;

[0053] a target statistic determination module, configured to determine a statistic value corresponding to each clustering result using a modeling model with Gaussian mixture model parameters, determine a target monitoring statistic by integrating all statistic values, and determine a target degree of freedom based on a variable of the target monitoring statistic;

[0054] An envelope baseline determination module is configured to search a target envelope baseline corresponding to the target degree of freedom and a preset confidence level from a preset database, wherein the database includes envelope baselines corresponding to different degrees of freedom and confidence levels;

[0055] A second acquisition module is configured to acquire current performance parameter test data of the component sample and determine a current statistic corresponding to the current performance parameter test data;

[0056] A first warning module is configured to, in response to the current statistic being greater than the target envelope baseline, determine that the component has a first type of failure risk, and issue a warning using a first warning rule preset corresponding to the first type of failure risk; or

[0057] The second warning module is configured to determine that the component has a second type of failure risk in response to the current statistic being less than or equal to the target envelope baseline, and issue a warning using a second warning rule preset corresponding to the second type of failure risk.

[0058] Based on the same inventive concept, the third aspect of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable by the processor, wherein the processor implements the method described in the first aspect above when executing the computer program.

[0059] Based on the same inventive concept, the fourth aspect of the present application provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute the method described in the first aspect above.

[0060] From the above, it can be seen that the component failure risk monitoring and early warning method and related equipment provided by the present application complete the classification of the performance parameter test data set of the component based on hierarchical clustering, cluster the variables with greater correlation into a cluster, and obtain multiple clustering results. Then, each clustering result is modeled based on the Gaussian mixture model, and then after the model is built, the envelope analysis statistic is calculated according to the definition of the statistic, and the target monitoring statistic is determined by combining the values ​​of the statistics of all clustering results, and then the target envelope baseline is determined based on this, and then the current performance parameter test data of the component sample is determined according to the comparison between the corresponding current statistic and the target envelope baseline. The type of failure risk is determined and warned, and the parameters of the components with the first type of failure risk can be traced, and the comprehensive monitoring of the mean, variance value and correlation coefficient between multiple variables can be realized, so that components with potential risks can be found in time, thereby ensuring the improvement of the quality and reliability of components. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions in the present application or related technologies, the drawings required for use in the embodiments or related technical descriptions are briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0062] Figure 1 A flowchart of a component failure risk monitoring and early warning method according to an embodiment of the present application;

[0063] Figure 2A A schematic diagram of a component failure risk monitoring and early warning model processing flow diagram of an embodiment of the present application;

[0064] Figure 2B A schematic diagram of the online monitoring, early warning and component parameter tracing process of an embodiment of the present application;

[0065] Figure 2C This is a schematic diagram of parameter distribution between years in the embodiment of the present application;

[0066] Figure 2D It is a characteristic schematic diagram that component data of different production batches of the embodiment of the present application still conforms to the Gaussian distribution;

[0067] Figure 2E A schematic diagram of a clustering tree after hierarchical clustering according to an embodiment of the present application;

[0068] Figure 2F A schematic diagram of component risk determination in an embodiment of the present application;

[0069] Figure 2G A first schematic diagram of parameters of failure risk according to an embodiment of the present application;

[0070] Figure 2H A second schematic diagram of parameters of failure risk according to an embodiment of the present application;

[0071] Figure 3 This is a structural block diagram of a component failure risk monitoring and early warning device according to an embodiment of the present application;

[0072] Figure 4 A schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0073] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0074] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present application should be the usual meanings understood by people with ordinary skills in the field to which the present application belongs. The "first", "second" and similar words used in the embodiments of the present application do not represent any order, quantity or importance, but are only used to distinguish different components. "Including" or "comprising" and similar words mean that the elements or objects appearing in front of the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0075] It is understandable that before using the technical solutions of each embodiment of the present application, the type, scope of use, usage scenarios, etc. of the personal information involved will be informed to the user in an appropriate manner, and the user's authorization will be obtained.

[0076] For example, in response to receiving an active request from a user, a prompt message is sent to the user to clearly remind the user that the operation requested to be performed will require obtaining and using the user's personal information. Thus, the user can independently choose whether to provide personal information to the electronic device, application, server, storage medium or other software or hardware that performs the operation of the technical solution of the present application according to the prompt message.

[0077] As an optional but non-limiting implementation, in response to receiving the user's active request, the prompt information may be sent to the user in the form of a pop-up window, in which the prompt information may be presented in text form. In addition, the pop-up window may also carry a selection control for the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0078] It is understandable that the above notification and the process of obtaining user authorization are merely illustrative and do not constitute a limitation on the implementation method of this application. Other methods that meet relevant laws and regulations may also be applied to the implementation method of this application.

[0079] Under the current background of accelerating the construction of equipment independent guarantee, large-scale batch production capacity and quality control point forward, equipment quality management faces new challenges. With the rapid development of large-scale automated production in the field of components and the application of advanced information technology, the number of product deliveries has increased exponentially, and the component quality assurance model based on post-management can no longer meet the needs of large-scale batch production. Therefore, it is urgent to break through the traditional quality reliability control method of "all inspections should be carried out", shift the focus of quality assurance work to supplier quality control and problem prevention, solidify quality requirements in component production and process links, strictly control from the source, and continuously promote the transformation and upgrading of component quality assurance model.

[0080] Currently, components are screened before leaving the factory. This process accumulates a large amount of performance parameter test data, which contains a lot of valuable information. However, there is currently a lack of effective intelligent analysis methods, resulting in the value of existing data resources not being fully utilized, and unable to provide data support for suppliers to carry out subsequent targeted component quality assurance work.

[0081] The embodiment of the present application provides a component failure risk monitoring and early warning method, which completes the classification of the performance parameter test data set of the component based on hierarchical clustering, clusters the variables with greater correlation into a cluster, and obtains multiple clustering results. Then, each clustering result is modeled based on the Gaussian mixture model, and then after the model is built, the envelope analysis statistic is calculated according to the definition of the statistic, and the target monitoring statistic is determined by combining the values ​​of the statistics of all clustering results, and then the target envelope baseline is determined based on this, and then the current performance parameter test data of the component sample is compared with the corresponding current statistic and the target envelope baseline to determine the type of failure risk and issue an early warning, and the parameters of the components with the first type of failure risk can be traced, and the comprehensive monitoring of the mean, variance value and correlation coefficient of multiple variables can be realized, so that components with potential risks can be discovered in time, thereby ensuring the improvement of the quality and reliability of components.

[0082] like Figure 1 As shown, the method of this embodiment includes:

[0083] Step 101, obtaining a performance parameter test data set of components in a normal state.

[0084] In this step, the component failure risk monitoring and early warning method of the embodiment of the present application is applicable to various types of components, such as semiconductor chips, capacitors, resistors, transistors, etc., which are not limited here.

[0085] Components refer to components that need to be monitored for failure risk. Normally, components are in normal working condition.

[0086] First, obtain the performance parameter test data set of components under normal conditions, which contains various performance parameter data generated by components under normal conditions. Among them, performance parameter data refers to various values ​​reflecting the performance of components, such as voltage, current, temperature, etc. These data can be collected in real time during the actual use of components, or can be queried from historical records.

[0087] In addition, the working data of components can be collected through various terminal devices and uploaded to the cloud server for centralized management. In addition, the collection cycle can be set according to actual needs, such as once per minute, once per hour, etc., and the number of collections per time can also be set according to actual needs, such as 10, 100, etc.

[0088] Step 102: perform hierarchical clustering on the performance parameter test data set to obtain multiple clustering results.

[0089] In this step, each performance parameter test data in the performance parameter test data set is first used as an initial cluster, and then the similarity between the initial clusters is compared two by two. If a certain similarity is met, they are merged into a new cluster, so that a new cluster set is obtained. Then, the clusters in the new cluster set are compared two by two for similarity, and this cycle is repeated until all the initial clusters are merged into the same large cluster, thereby obtaining a nested clustering tree. Finally, multiple clustering results are obtained based on the nested clustering tree. In this way, a large amount of performance parameter test data can be quickly and accurately classified, which is convenient for subsequent analysis and processing.

[0090] Step 103, modeling each clustering result based on the Gaussian mixture model to obtain a modeling model, and determining the Gaussian mixture model parameters of the modeling model by using the maximum expectation algorithm.

[0091] In this step, the performance parameter test data in each clustering result are first initialized, including the initialization of the mean, variance value and prior probability. Then, the maximum expectation algorithm is used to iteratively update the mean, variance value and prior probability of the performance parameter test data in each clustering result until the stopping criterion is met. Finally, the updated mean, variance value and prior probability are used as the parameters of the Gaussian mixture model.

[0092] According to the Gaussian mixture model parameters, a Gaussian mixture model corresponding to each clustering result can be constructed, that is, the performance parameter test data in each clustering result is approximately expressed by a set of mixed Gaussian distributions, so as to better characterize their distribution characteristics. In addition, the parameters of the Gaussian mixture model can be adjusted by fitting it to make it more in line with the actual situation.

[0093] Step 104, using a modeling model with Gaussian mixture model parameters to determine the statistical value corresponding to each clustering result, combining all statistical values ​​to determine the target monitoring statistic, and determining the target degrees of freedom based on the variables of the target monitoring statistic.

[0094] In this step, for each clustering result, the corresponding Gaussian mixture model can be used to obtain the statistics corresponding to each sample in the clustering result, and then the statistics of the corresponding samples in all clustering results are combined to form the target monitoring statistics, which can be expressed as follows:

[0095] d m =d 1m +d 2m +......+d Hm

[0096] Among them, H represents the Hth clustering result, m represents the mth sample in the corresponding clustering result, and d m Represents the target monitoring statistic corresponding to the mth sample.

[0097] Next, we can determine the target degrees of freedom based on this target monitoring statistic. Specifically, we can use the chi-square distribution method to determine the target degrees of freedom. From the above, we can see that if the target monitoring statistic obeys the chi-square distribution with a certain degree of freedom, then this degree of freedom is the target degree of freedom.

[0098] Step 105 , searching a preset database for a target envelope baseline corresponding to the target degree of freedom and a preset confidence level, wherein the database includes envelope baselines corresponding to different degrees of freedom and confidence levels.

[0099] In this step, a database is pre-established, which includes envelope baselines corresponding to different degrees of freedom and confidence levels.

[0100] On this basis, the envelope baseline corresponding to the target degrees of freedom and confidence can be quickly found.

[0101] Step 106, obtaining current performance parameter test data of the component sample, and determining current statistics corresponding to the current performance parameter test data.

[0102] In this step, the current performance parameter test data of the component sample is obtained. Then, the current performance parameter test data of the component sample is statistically processed to obtain a statistic. This statistic can reflect the overall performance of the component, such as the average value, standard deviation, etc.

[0103] Among them, the current performance parameter test data of different components They often have different dimensions and units, and the original data ranges are generally different, which may have an adverse effect on the convergence speed and accuracy of data analysis and mathematical modeling. Therefore, it is necessary to normalize them and map the values ​​to a unified range of [0, 1] to ensure that the parameters are compared and analyzed at the same level, and obtain normalized data.

[0104] Then, according to the offline hierarchical clustering results, different Gaussian mixture models are called to calculate Monitoring the value of a statistic (i.e. current statistics).

[0105] Step 107: In response to the current statistic being greater than the target envelope baseline, it is determined that the component has a first type of failure risk, and a first warning rule corresponding to the first type of failure risk is used to issue a warning. Or,

[0106] In this step, if the current statistic is greater than the target envelope baseline, it means that the performance of the component at the current moment exceeds expectations, and there may be a first type of failure risk (indicating a high risk), and a first warning rule corresponding to the first type of failure risk is used for warning. At this time, corresponding measures need to be taken to avoid the occurrence of failures, such as increasing the monitoring frequency, replacing parts, etc.

[0107] Among them, the first warning rule can be an audible and visual alarm.

[0108] Step 108 , in response to the current statistic being less than or equal to the target envelope baseline, determining that the component has a second type of failure risk, and issuing an early warning using a second early warning rule preset corresponding to the second type of failure risk.

[0109] In this step, on the contrary, if the current statistic is less than or equal to the target envelope baseline, it means that the performance of the component at the current moment has a second type of failure risk (indicating a lower risk), and a warning is issued using the second warning rule preset corresponding to the second type of failure risk.

[0110] Among them, the second warning rule can be an information prompt alarm.

[0111] Through the above scheme, the performance parameter test data set of components is classified based on hierarchical clustering, and the variables with large correlation are clustered into a cluster to obtain multiple clustering results. Then, each clustering result is modeled based on the Gaussian mixture model, and then after the model is built, the envelope analysis statistic is calculated according to the definition of the statistic, and the target monitoring statistic is determined by combining the values ​​of the statistics of all clustering results, and then the target envelope baseline is determined based on this. Then, the current performance parameter test data of the component sample is compared with the corresponding current statistic and the target envelope baseline to determine the type of failure risk and issue an early warning. The parameters of the components with the first type of failure risk can be traced, and the comprehensive monitoring of the mean, variance value and correlation coefficient of multiple variables can be realized. Components with potential risks can be discovered in time, thereby ensuring the improvement of component quality and reliability.

[0112] In some embodiments, the current statistic includes a plurality of current sub-statistics respectively corresponding to current performance parameter test data.

[0113] After step 107, the method further includes:

[0114] Step A1, performing the following operations on each of the sub-statistics:

[0115] Step A11, determining a chi-square distribution density function satisfied by the sub-statistic;

[0116] Step A12, integrating the chi-square distribution density function to obtain the risk probability of the current performance parameter test data corresponding to the sub-statistic;

[0117] Step A2, sorting the risk probabilities of the current performance parameter test data according to a preset sorting rule to obtain a sorting result.

[0118] Step A3: selecting a predetermined number of current performance parameter test data from the sorting results as failure risk parameters of the components.

[0119] In the above scheme, when it is determined that a component has a risk of failure, parameter traceability analysis is performed on the component with the risk of failure, the risk probability of each current performance parameter test data is calculated, and the top three current performance parameter test data with the highest risk probability are output as the result. The experimental results on a certain model specification data set of components show that the proposed envelope analysis method can effectively discover the potential risks of components, quickly locate the current performance parameter test data of the component that causes the problem, discover the quality risks that may cause accidents in advance, and ensure the improvement of the quality reliability of components.

[0120] Among them, the value of risk probability P-value is determined by the following process:

[0121] The P-value is obtained by integrating the chi-square distribution density function in the range [0, F]. x represents the Gaussian mixture estimate of a single parameter, which is calculated using the standard mean and variance of the parameters corresponding to the model training. The chi-square distribution density function is as follows:

[0122]

[0123] In addition, the preset sorting rule can be sorting from high to bottom or from low to high. The present application prefers sorting from high to bottom.

[0124] In some embodiments, the performance parameter test data set includes a plurality of performance parameter test data.

[0125] Step 102 includes:

[0126] Step B1: taking each performance parameter test data as an initial cluster, performing Pearson correlation coefficient processing on each of the multiple initial clusters to obtain multiple first Pearson correlation coefficients.

[0127] Step B2, counting the largest first Pearson correlation coefficient among the multiple first Pearson correlation coefficients, merging the two initial clusters corresponding to the largest first Pearson correlation coefficient to obtain a first merged cluster, and using the first merged cluster as a new cluster.

[0128] Step B3: performing a cluster merging process on each of the other clusters in the plurality of initial clusters except the two clusters merged into the new cluster and the new cluster. The cluster merging process is as follows:

[0129] Step B4, performing Pearson correlation coefficient processing on each of the other clusters and the new clusters, to obtain a plurality of second Pearson correlation coefficients.

[0130] Step B5, counting the largest second Pearson correlation coefficient among the second Pearson correlation coefficients, merging the two clusters corresponding to the largest second Pearson correlation coefficient to obtain a second merged cluster, using the second merged cluster as the new cluster, repeating the cluster merging process until all initial clusters are merged into the same cluster or the preset clustering termination condition is reached, obtaining a nested clustering tree, and obtaining the multiple clustering results based on the nested clustering tree.

[0131] In the above scheme, 1) different performance parameter test data often have different dimensions and units, and the original data ranges are generally different, which may have an adverse effect on the convergence speed and accuracy of data analysis and mathematical modeling. Therefore, it is necessary to normalize and preprocess the data to map the values ​​into a unified range of [0, 1] to ensure that each parameter is compared and analyzed at the same level.

[0132] 2) Consider each performance parameter test data of the component as an initial cluster, then there are N initial clusters in total, and calculate the distance between different clusters based on the Pearson correlation coefficient. The larger the value of the correlation coefficient between two clusters, the closer the distance between the two clusters.

[0133] 3) Find the two clusters with the closest distance and merge them to produce a new cluster.

[0134] 4) Recalculate the distance between the newly generated cluster and the original old cluster, and merge the two clusters with the closest distance again.

[0135] 5) Repeat the above process until all clusters are merged into one cluster or a termination condition is reached (such as reaching a preset number of repeated iterations), and a nested clustering tree is obtained, and multiple clustering results are obtained based on the nested clustering tree.

[0136] Hierarchical clustering does not require the number of clusters to be specified in advance. This feature makes hierarchical clustering more flexible when processing data and can form clusters based on the natural distribution of data.

[0137] In some embodiments, step 103 includes:

[0138] Step C1, perform the following operations on each clustering result:

[0139] Step C11, initializing the mean, variance value and prior probability of each Gaussian component.

[0140] Step C12, determining, in an initialized state, a first posterior probability that each performance parameter test data in the clustering result belongs to each Gaussian component.

[0141] Step C13: updating the mean, variance value and prior probability of each Gaussian component by maximizing the likelihood function based on the first posterior probability.

[0142] Step C14, taking the updated mean of each Gaussian component as the target mean, taking the updated variance value of each Gaussian component as the target variance, and taking the updated prior probability of each Gaussian component as the target prior probability, and performing an update iterative process for the target mean, target variance value, and target prior probability corresponding to each Gaussian component. The update iterative process is as follows:

[0143] Step C15, determining in an updated state the second posterior probability that each performance parameter test data in the clustering result belongs to each Gaussian component.

[0144] Step C16, based on the second posterior probability, the mean, variance value and prior probability of each Gaussian component are updated by maximizing the likelihood function, the updated mean of each Gaussian component is used as the target mean, the updated variance value of each Gaussian component is used as the target variance value, and the updated prior probability of each Gaussian component is used as the target prior probability, and the update iterative process is repeated for the target mean, target variance value and target prior probability corresponding to each Gaussian component until the preset update iterative termination condition is reached, and the final mean, final variance value and final prior probability corresponding to each Gaussian component are obtained.

[0145] Step C17, taking the final mean, final variance value and final prior probability corresponding to each Gaussian component as the Gaussian mixture model parameters.

[0146] In the above scheme, the probability distribution of the Gaussian mixture model is:

[0147]

[0148] Among them, x represents the test data of each performance parameter in the corresponding clustering result, K represents the number of sub-Gaussian models (Gaussian components) in the Gaussian mixture model, and α k is the prior that x belongs to the kth sub-Gaussian model, is the Gaussian distribution density function of the kth sub-Gaussian model, θ k (μ k , σ k 2 ).

[0149] For the Gaussian mixture model, the parameters It is the expectation, variance value (or covariance value) of each sub-Gaussian model and the probability of occurrence in the Gaussian mixture model. The steps of using the Gaussian mixture model to analyze and model the performance parameter test data are as follows:

[0150] 1) Starting from the first clustering result, the normalized data is used as the input of the Gaussian mixture model, the number of components of the Gaussian mixture model is set to T, and the model parameters θ are randomly initialized. (0) .

[0151] 2) Use the Expectation Maximization (EM) algorithm to estimate the parameters of the model. The M step (maximization step) is to re-estimate the model parameters using maximum likelihood and update the parameters for the next iterative E step (expectation step).

[0152] 3) Repeat steps E and M until the predetermined stop condition is reached (parameter convergence or the maximum number of iterations is reached), and the final Gaussian mixture model parameter θ is obtained. (1) .

[0153] In some embodiments, in step 104, the modeling model with Gaussian mixture model parameters is obtained by the following process:

[0154] Step D1, perform the following operations on each clustering result:

[0155] Step D11 : for each Gaussian component, determine the target Gaussian component parameters based on the corresponding final mean value and final variance value.

[0156] Step D12, integrating the target Gaussian component parameters using the corresponding final priori probabilities to obtain the Gaussian mixture model.

[0157] In the above scheme, the probability distribution of the Gaussian mixture model is:

[0158]

[0159] Among them, x represents the test data of each performance parameter in the corresponding clustering result, K represents the number of sub-Gaussian models (Gaussian components) in the Gaussian mixture model, and α k is the prior probability that x belongs to the kth sub-Gaussian model, is the Gaussian distribution density function of the kth sub-Gaussian model, θ k =(μ k , σ k 2 ).

[0160] For the Gaussian mixture model, the parameters It is the expectation, variance value (or covariance value) of each sub-Gaussian model and the probability of occurrence in the Gaussian mixture model.

[0161] For each clustering result, after determining the target Gaussian component parameters of each corresponding Gaussian component, the corresponding final prior probability is used to integrate the target Gaussian component parameters, so as to quickly obtain the Gaussian mixture model corresponding to the clustering result.

[0162] In some embodiments, the modeling model with Gaussian mixture model parameters includes multiple Gaussian components, and each clustering result includes multiple performance parameter test data.

[0163] In step 104, the statistical value corresponding to each clustering result is determined by using the modeling model with Gaussian mixture model parameters, and the target monitoring statistic is determined by integrating all the statistical values, including:

[0164] Step E1: taking any performance parameter test data among the plurality of performance parameter test data as target performance parameter test data, and performing the following operations on the target performance parameter test data in each clustering result:

[0165] Step E11, determining the distance between the target performance parameter test data and the center of each Gaussian component, and summing up each distance to obtain an initial statistic corresponding to the target performance parameter test data.

[0166] Step E2, summing up the initial statistics corresponding to the target performance parameter test data in each clustering result to obtain the target monitoring statistic.

[0167] In the above scheme, the sum of the distances from the performance parameter test data of each component in each clustering result to the centers of T Gaussian components is calculated as the initial statistic, denoted as D′1={d 1m}(m=1, 2, 3, ..., M), and use the same method to calculate the initial statistics of each performance parameter test data in the remaining clustering results.

[0168] Then, by summing up the initial statistics of the corresponding samples in all clustering results, the target monitoring statistics can be quickly obtained, which can be expressed as follows:

[0169] d m =d 1m +d 2m +......+d Hm

[0170] Among them, H represents the Hth clustering result, m represents the mth sample in the corresponding clustering result, and d m Represents the target monitoring statistic corresponding to the mth sample.

[0171] In some embodiments, step 101 includes:

[0172] Step F1, obtaining a plurality of initial performance parameter test data of the component in a normal state.

[0173] Step F2, normalizing the multiple initial performance parameter test data to obtain the performance parameter test data set.

[0174] In the above scheme, different initial performance parameter test data often have different dimensions and units, and there are generally differences in the original data range, which may have an adverse effect on the convergence speed and accuracy of data analysis and mathematical modeling. Therefore, it is necessary to normalize and preprocess the data to map the values ​​into a unified range of [0, 1] to obtain a performance parameter test data set to ensure that each parameter is compared and analyzed at the same level.

[0175] In some embodiments, the component failure risk monitoring and early warning model of the present application includes three parts: hierarchical clustering, Gaussian mixture estimation, and early warning baseline formation. Figure 2A As shown:

[0176] Step 2011, preprocessing of initial performance parameter test data set of components;

[0177] Step 2012, hierarchical clustering;

[0178] Step 213, Gaussian mixture estimation, for each cluster (i.e., clustering result), obtain the final Gaussian mixture model parameters, such as the mean, covariance value, and prior probability corresponding to cluster 1, the mean, covariance value, and prior probability corresponding to cluster 2, and the mean, covariance value, and prior probability corresponding to cluster N;

[0179] Step 2014, calculate the statistics corresponding to each variable in each cluster (i.e., the initial statistics). For example, the statistics corresponding to a sample in cluster 1 is D1, the statistics corresponding to the corresponding sample in cluster 2 is D2, and the statistics corresponding to the corresponding sample in cluster N is D N

[0180] Step 2015, obtaining a statistic D (i.e., a target monitoring statistic) based on the sum of the statistics of the corresponding variables in each cluster;

[0181] Step 2016, setting the confidence level;

[0182] Step 2017, forming an envelope warning baseline (ie, a target envelope baseline).

[0183] In addition, the online monitoring and early warning and component parameter traceability process of this application includes three parts: online monitoring and early warning, risk probability calculation and risk parameter determination, such as Figure 2B As shown:

[0184] Step 2021, input component parameter detection data;

[0185] Step 2022, calculating the value of the statistic;

[0186] Step 2023, determine whether it exceeds the envelope baseline, if not, return to step 2021, if yes, execute step 2024;

[0187] Step 2024: tracing the parameters to obtain the risk probability of each parameter, such as the risk probability of parameter 1 P1, ..., the risk probability of parameter K P K ;

[0188] Step 2025, sorting parameters according to risk probability;

[0189] Step 2026, obtaining the top three riskiest parameters.

[0190] The effect verification process of this application is as follows:

[0191] After processing, the historical data of a certain model and specification component has a total of 5436 data, a total of 41 test parameters, and the production batches of the data are shown in Table 1 below:

[0192] Table 1

[0193] Batch No. Batch situation Number of failure analysis records 1 normal 0 2 There are failed devices 2 3 normal 0 4 normal 0 5 normal 0 6 normal 0

[0194] (1) Parameter distribution analysis

[0195] For parameters of the same type with multiple pins, the attribute dimension distribution diagram basically follows the Gaussian distribution. Taking parameters: iih_a1, iih_a2, iih_a3, iih_a4, ioz_y1, ioz_y2, ioz_y3, ioz_y4 as an example, it can be seen that the parameter distribution varies greatly between years, such as Figure 2C shown.

[0196] The component data of different production batches still conform to the characteristics of Gaussian distribution, such as Figure 2D shown.

[0197] Taking the same type of parameters cont and iih as an example, we can see that the distributions of cont_g_G, cont_a1, cont_a2, and cont_a3 are similar, while the distributions of the same type of parameters iih_g_G, iih_a1, iih_a2, and iih_a3 are quite different.

[0198] (2) Hierarchical cluster analysis

[0199] There are 41 total test parameters for this model specification. After hierarchical clustering, the test parameters form 15 parameter clusters, such as Figure 2E shown.

[0200] (3) Component risk assessment

[0201] First, we train the 1st, 3rd, 4th, 5th, and 6th batches of qualified devices to obtain the envelope baseline of normal data. When the value of the statistic exceeds this line, it indicates that the component is at risk of failure. We input the test set data into the model using the data from the second batch as an example, and obtain Figure 2F of the test results.

[0202] from Figure 2F It can be seen that a total of 2 sample points (components 1 and 3) exceed the control line (i.e., the envelope baseline), indicating that two components in the second batch have the first type of failure risk.

[0203] (4) Component parameter traceability

[0204] There are two components in Batch 2 with Type I failure risk. The parameter traceability results are as follows:

[0205] The parameters that cause component 1 to fail are mainly iih_a2_A2, cont_a4_A4, and ios_y_0_y4_Y4, as follows Figure 2G shown.

[0206] The parameters that cause component 2 to have failure risk are mainly cont_a2_A2, cont_a3_A3, and vic_a2_A2. Figure 2H shown.

[0207] It should be noted that the method of the embodiment of the present application can be performed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiment of the present application, and the multiple devices will interact with each other to complete the described method.

[0208] It should be noted that the above describes some embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the above embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0209] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present application also provides a component failure risk monitoring and early warning device.

[0210] refer to Figure 3 , the component failure risk monitoring and early warning device comprises:

[0211] The first acquisition module 301 is configured to acquire a performance parameter test data set of a component in a normal state;

[0212] The clustering module 302 is configured to perform hierarchical clustering on the performance parameter test data set to obtain multiple clustering results;

[0213] The parameter determination module 303 is configured to model each clustering result based on the Gaussian mixture model to obtain a modeling model, and determine the Gaussian mixture model parameters of the modeling model by using the maximum expectation algorithm;

[0214] A target statistic determination module 304 is configured to determine a statistic value corresponding to each clustering result using a modeling model with Gaussian mixture model parameters, determine a target monitoring statistic by integrating all the statistic values, and determine a target degree of freedom based on a variable of the target monitoring statistic;

[0215] The envelope baseline determination module 305 is configured to search a target envelope baseline corresponding to the target degree of freedom and a preset confidence level from a preset database, wherein the database includes envelope baselines corresponding to different degrees of freedom and confidence levels;

[0216] The second acquisition module 306 is configured to acquire current performance parameter test data of the component sample and determine a current statistic corresponding to the current performance parameter test data;

[0217] The first warning module 307 is configured to, in response to the current statistic being greater than the target envelope baseline, determine that the component has a first type of failure risk, and issue a warning using a first warning rule preset corresponding to the first type of failure risk; or

[0218] The second warning module 308 is configured to determine that the component has a second type of failure risk in response to the current statistic being less than or equal to the target envelope baseline, and issue a warning using a second warning rule preset corresponding to the second type of failure risk.

[0219] In some embodiments, the current statistic includes a plurality of current sub-statistics corresponding to current performance parameter test data respectively;

[0220] The component failure risk monitoring and early warning device further includes a risk parameter determination module. In response to the current statistic being greater than the target envelope baseline, after determining that the component has a first type of failure risk, the risk parameter determination module is specifically configured as follows:

[0221] For each of the sub-statistics described the following operations are performed:

[0222] Determining a chi-square distribution density function satisfied by the sub-statistic;

[0223] Integrating the chi-square distribution density function to obtain the risk probability of the current performance parameter test data corresponding to the sub-statistic;

[0224] Sort the risk probabilities of the current performance parameter test data according to the preset sorting rules to obtain the sorting results;

[0225] A predetermined number of current performance parameter test data are selected from the sorting results as failure risk parameters of the components.

[0226] In some embodiments, the performance parameter test data set includes a plurality of performance parameter test data;

[0227] The clustering module 302 is specifically configured as follows:

[0228] Taking each performance parameter test data as an initial cluster, performing Pearson correlation coefficient processing on each of the multiple initial clusters to obtain multiple first Pearson correlation coefficients;

[0229] Counting the largest first Pearson correlation coefficient among the multiple first Pearson correlation coefficients, merging two initial clusters corresponding to the largest first Pearson correlation coefficient to obtain a first merged cluster, and using the first merged cluster as a new cluster;

[0230] A cluster merging process is performed on each of the other clusters in the multiple initial clusters except the two clusters merged into the new cluster and the new cluster. The cluster merging process is as follows:

[0231] Performing Pearson correlation coefficient processing on each of the other clusters and the new clusters, to obtain a plurality of second Pearson correlation coefficients;

[0232] The largest second Pearson correlation coefficient among the second Pearson correlation coefficients is counted, and the two clusters corresponding to the largest second Pearson correlation coefficient are merged to obtain a second merged cluster. The second merged cluster is used as the new cluster, and the cluster merging process is repeated until all initial clusters are merged into the same cluster or a preset clustering termination condition is reached to obtain a nested clustering tree, and the multiple clustering results are obtained based on the nested clustering tree.

[0233] In some embodiments, the parameter determination module 303 is specifically configured to:

[0234] For each clustering result, the following operations are performed:

[0235] Initialize the mean, variance and prior probability of each Gaussian component;

[0236] Determine, in an initialization state, a first posterior probability that each performance parameter test data in the clustering result belongs to each Gaussian component;

[0237] Based on the first posterior probability, the mean, variance value and prior probability of each Gaussian component are updated by maximizing the likelihood function;

[0238] The updated mean of each Gaussian component is used as the target mean, the updated variance value of each Gaussian component is used as the target variance value, and the updated prior probability of each Gaussian component is used as the target prior probability. An update iteration process is performed for the target mean, target variance value, and target prior probability corresponding to each Gaussian component. The update iteration process is as follows:

[0239] Determine in an updated state the second posterior probability that each performance parameter test data in the clustering result belongs to each Gaussian component;

[0240] Based on the second posterior probability, the mean, variance value and prior probability of each Gaussian component are updated by maximizing the likelihood function, the updated mean of each Gaussian component is used as the target mean, the updated variance value of each Gaussian component is used as the target variance value, and the updated prior probability of each Gaussian component is used as the target prior probability, and the update iteration process is repeatedly performed for the target mean, target variance value and target prior probability corresponding to each Gaussian component until a preset update iteration termination condition is reached, so as to obtain the final mean, final variance value and final prior probability corresponding to each Gaussian component;

[0241] The final mean, final variance value and final prior probability corresponding to each Gaussian component are used as the Gaussian mixture model parameters.

[0242] In some embodiments, the target statistic determination module 304 is specifically configured to:

[0243] For each clustering result, the following operations are performed:

[0244] For each Gaussian component, determining a target Gaussian component parameter based on the corresponding final mean value and final variance value;

[0245] Based on the parameters of each target Gaussian component, the corresponding final prior probability is used for integration to obtain the modeling model with Gaussian mixture model parameters.

[0246] In some embodiments, the modeling model with Gaussian mixture model parameters includes multiple Gaussian components, and each clustering result includes multiple performance parameter test data;

[0247] The target statistic determination module 304 is specifically configured to:

[0248] Any performance parameter test data among the plurality of performance parameter test data is used as target performance parameter test data, and the following operations are performed for each target performance parameter test data in the clustering result:

[0249] Determine the distance between the target performance parameter test data and the center of each Gaussian component, and sum each distance to obtain an initial statistic corresponding to the target performance parameter test data;

[0250] The initial statistics corresponding to the target performance parameter test data in each clustering result are summed to obtain the target monitoring statistics.

[0251] In some embodiments, the first acquisition module 301 is specifically configured to:

[0252] Acquire a plurality of initial performance parameter test data of the components in a normal state;

[0253] The multiple initial performance parameter test data are normalized to obtain the performance parameter test data set.

[0254] For the convenience of description, the above device is described in terms of functions divided into various modules. Of course, when implementing the present application, the functions of each module can be implemented in the same or multiple software and / or hardware.

[0255] The device of the above embodiment is used to implement the corresponding component failure risk monitoring and early warning method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0256] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments, the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the component failure risk monitoring and early warning method described in any of the above embodiments is implemented.

[0257] Figure 4 A more specific schematic diagram of the hardware structure of an electronic device provided in this embodiment is shown, and the device may include: a processor 401, a memory 402, an input / output interface 403, a communication interface 404, and a bus 405. The processor 401, the memory 402, the input / output interface 403, and the communication interface 404 are connected to each other in communication within the device through the bus 405.

[0258] The processor 401 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0259] The memory 402 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 402 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program codes are stored in the memory 402 and called and executed by the processor 401.

[0260] The input / output interface 403 is used to connect the input / output module to realize information input and output. The input / output module can be configured in the device as a component (not shown in the figure), or it can be externally connected to the device to provide corresponding functions. The input device may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device may include a display, a speaker, a vibrator, an indicator light, etc.

[0261] The communication interface 404 is used to connect a communication module (not shown) to realize communication interaction between the device and other devices. The communication module can realize communication through a wired mode (such as USB, network cable, etc.) or a wireless mode (such as mobile network, WIFI, Bluetooth, etc.).

[0262] The bus 405 comprises a pathway for transmitting information between the various components of the device (eg, the processor 401 , the memory 402 , the input / output interface 403 , and the communication interface 404 ).

[0263] It should be noted that, although the above device only shows the processor 401, the memory 402, the input / output interface 403, the communication interface 404 and the bus 405, in the specific implementation process, the device may also include other components necessary for normal operation. In addition, it can be understood by those skilled in the art that the above device may also only include the components necessary for implementing the embodiments of the present specification, and does not necessarily include all the components shown in the figure.

[0264] The electronic device of the above embodiment is used to implement the corresponding component failure risk monitoring and early warning method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0265] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present application also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the component failure risk monitoring and early warning method as described in any of the above embodiments.

[0266] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0267] The computer instructions stored in the storage medium of the above embodiment are used to enable the computer to execute the component failure risk monitoring and early warning method described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0268] A person skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present application is limited to these examples. In line with the concept of the present application, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present application as described above, which are not provided in detail for the sake of simplicity.

[0269] In addition, to simplify the description and discussion, and in order not to make the embodiments of the present application difficult to understand, the known power supply / ground connection with the integrated circuit (IC) chip and other components may or may not be shown in the provided drawings. In addition, the device can be shown in the form of a block diagram to avoid making the embodiments of the present application difficult to understand, and this also takes into account the fact that the details of the implementation of these block diagram devices are highly dependent on the platform to be implemented in the embodiments of the present application (that is, these details should be fully within the scope of understanding of those skilled in the art). In the case of elaborating specific details (e.g., circuits) to describe exemplary embodiments of the present application, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without these specific details or when these specific details are changed. Therefore, these descriptions should be considered to be illustrative rather than restrictive.

[0270] Although the present application has been described in conjunction with specific embodiments of the present application, many replacements, modifications and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may use the embodiments discussed.

[0271] The embodiments of the present application are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the present application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present application should be included in the protection scope of the present application.

Claims

1. A component failure risk monitoring and early warning method, characterized in that: include: Obtain performance parameter test data sets of components under normal conditions; Performing hierarchical clustering on the performance parameter test data set to obtain multiple clustering results; Modeling each clustering result based on the Gaussian mixture model to obtain a modeling model, and determining the Gaussian mixture model parameters of the modeling model through the maximum expectation algorithm; Determine a statistical value corresponding to each clustering result using a modeling model with Gaussian mixture model parameters, determine a target monitoring statistic by integrating all statistical values, and determine a target degree of freedom based on a variable of the target monitoring statistic; Searching for a target envelope baseline corresponding to the target degree of freedom and a preset confidence level from a preset database, wherein the database includes envelope baselines corresponding to different degrees of freedom and confidence levels; Acquire current performance parameter test data of the component sample, and determine current statistics corresponding to the current performance parameter test data; In response to the current statistic being greater than the target envelope baseline, determining that the component has a first type of failure risk, and issuing an early warning using a first early warning rule preset corresponding to the first type of failure risk; or, In response to the current statistic being less than or equal to the target envelope baseline, it is determined that the component has a second type of failure risk, and a warning is issued using a second warning rule preset corresponding to the second type of failure risk.

2. The method according to claim 1, characterized in that: The current statistic includes a plurality of current sub-statistics respectively corresponding to the current performance parameter test data; In response to the current statistic being greater than the target envelope baseline, after determining that the component has a first type of failure risk, the method includes: For each of the sub-statistics described the following operations are performed: Determining a chi-square distribution density function satisfied by the sub-statistic; Integrating the chi-square distribution density function to obtain the risk probability of the current performance parameter test data corresponding to the sub-statistic; Sort the risk probabilities of the current performance parameter test data according to the preset sorting rules to obtain the sorting results; A predetermined number of current performance parameter test data are selected from the sorting results as failure risk parameters of the components.

3. The method according to claim 1, characterized in that The performance parameter test data set includes a plurality of performance parameter test data; The hierarchical clustering is performed on the performance parameter test data set to obtain multiple clustering results, including: Taking each performance parameter test data as an initial cluster, performing Pearson correlation coefficient processing on each of the multiple initial clusters to obtain multiple first Pearson correlation coefficients; Counting the largest first Pearson correlation coefficient among the multiple first Pearson correlation coefficients, merging two initial clusters corresponding to the largest first Pearson correlation coefficient to obtain a first merged cluster, and using the first merged cluster as a new cluster; A cluster merging process is performed on each of the other clusters in the multiple initial clusters except the two clusters merged into the new cluster and the new cluster. The cluster merging process is as follows: Performing Pearson correlation coefficient processing on each of the other clusters and the new clusters, to obtain a plurality of second Pearson correlation coefficients; The largest second Pearson correlation coefficient among the second Pearson correlation coefficients is counted, and the two clusters corresponding to the largest second Pearson correlation coefficient are merged to obtain a second merged cluster. The second merged cluster is used as the new cluster, and the cluster merging process is repeated until all initial clusters are merged into the same cluster or a preset clustering termination condition is reached to obtain a nested clustering tree, and the multiple clustering results are obtained based on the nested clustering tree.

4. The method according to claim 1, characterized in that: The method of determining the Gaussian mixture model parameters of the modeling model by using the maximum expectation algorithm includes: For each clustering result, the following operations are performed: Initialize the mean, variance and prior probability of each Gaussian component; Determine, in an initialization state, a first posterior probability that each performance parameter test data in the clustering result belongs to each Gaussian component; Based on the first posterior probability, the mean, variance value and prior probability of each Gaussian component are updated by maximizing the likelihood function; The updated mean of each Gaussian component is used as the target mean, the updated variance value of each Gaussian component is used as the target variance value, and the updated prior probability of each Gaussian component is used as the target prior probability. An update iteration process is performed for the target mean, target variance value, and target prior probability corresponding to each Gaussian component. The update iteration process is as follows: Determine in an updated state the second posterior probability that each performance parameter test data in the clustering result belongs to each Gaussian component; Based on the second posterior probability, the mean, variance value and prior probability of each Gaussian component are updated by maximizing the likelihood function, the updated mean of each Gaussian component is used as the target mean, the updated variance value of each Gaussian component is used as the target variance value, and the updated prior probability of each Gaussian component is used as the target prior probability, and the update iteration process is repeatedly performed for the target mean, target variance value and target prior probability corresponding to each Gaussian component until a preset update iteration termination condition is reached, so as to obtain the final mean, final variance value and final prior probability corresponding to each Gaussian component; The final mean, final variance value and final prior probability corresponding to each Gaussian component are used as the Gaussian mixture model parameters.

5. The method according to claim 4, characterized in that The modeling model with Gaussian mixture model parameters is obtained by the following process: For each clustering result, the following operations are performed: For each Gaussian component, determining a target Gaussian component parameter based on the corresponding final mean value and final variance value; Based on the parameters of each target Gaussian component, the corresponding final prior probability is used for integration to obtain the modeling model with Gaussian mixture model parameters.

6. The method according to claim 1, characterized in that The modeling model with Gaussian mixture model parameters includes multiple Gaussian components, and each clustering result includes multiple performance parameter test data; The method of using a modeling model with Gaussian mixture model parameters to determine a statistical value corresponding to each clustering result and combining all statistical values ​​to determine a target monitoring statistic includes: Any performance parameter test data among the plurality of performance parameter test data is used as target performance parameter test data, and the following operations are performed for each target performance parameter test data in the clustering result: Determine the distance between the target performance parameter test data and the center of each Gaussian component, and sum each distance to obtain an initial statistic corresponding to the target performance parameter test data; The initial statistics corresponding to the target performance parameter test data in each clustering result are summed to obtain the target monitoring statistics.

7. The method according to claim 1, characterized in that The step of obtaining a performance parameter test data set of a component in a normal state includes: Acquire a plurality of initial performance parameter test data of the components in a normal state; The multiple initial performance parameter test data are normalized to obtain the performance parameter test data set.

8. A component failure risk monitoring and early warning device, characterized in that: include: A first acquisition module is configured to acquire a performance parameter test data set of a component in a normal state; A clustering module is configured to perform hierarchical clustering on the performance parameter test data set to obtain multiple clustering results; A parameter determination module is configured to model each clustering result based on a Gaussian mixture model to obtain a modeling model, and determine Gaussian mixture model parameters of the modeling model by a maximum expectation algorithm; a target statistic determination module, configured to determine a statistic value corresponding to each clustering result using a modeling model with Gaussian mixture model parameters, determine a target monitoring statistic by integrating all statistic values, and determine a target degree of freedom based on a variable of the target monitoring statistic; An envelope baseline determination module is configured to search a target envelope baseline corresponding to the target degree of freedom and a preset confidence level from a preset database, wherein the database includes envelope baselines corresponding to different degrees of freedom and confidence levels; A second acquisition module is configured to acquire current performance parameter test data of the component sample and determine a current statistic corresponding to the current performance parameter test data; A first warning module is configured to, in response to the current statistic being greater than the target envelope baseline, determine that the component has a first type of failure risk, and issue a warning using a first warning rule preset corresponding to the first type of failure risk; or The second warning module is configured to determine that the component has a second type of failure risk in response to the current statistic being less than or equal to the target envelope baseline, and issue a warning using a second warning rule preset corresponding to the second type of failure risk.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to enable a computer to execute the method according to any one of claims 1 to 7.

Images